1. Field of the Invention
The present invention relates to a technique for encoding image information, particularly moving picture information, and a technique for decoding encoded moving picture information.
2. Description of the Related Art
In recent years, the needs for higher definition in image information have increased. Accordingly, television broadcasting is undergoing a transition from conventional SD (standard definition) with 720×480 pixels to HD (high definition) with 1920×1080 pixels.
The increase in data volume associated with higher definitions has led to demands for the development of a coding algorithm having higher efficiency. Efforts towards standardization of a compression coding scheme using interframe prediction are currently underway by ITU-T SG16 and ISO/IEC JTCl/SC29/WG11.
As it stands, the H.264/AVC (MPEG-4 PART10) encoding scheme is considered to have the highest encoding efficiency. One of the technical features newly introduced to this coding scheme allows selection of a reference picture to be used in interframe predictive coding from a plurality of frames. In other words, even a temporally-distant picture frame may be used as a reference picture if an improvement in coding efficiency may be expected.
However, the realization of high-efficiency coding by allowing such flexible selection of a reference picture presents a problem during special playback (refer to Japanese Patent Laid-Open No. 2004-328511 [reference document D1])
Conventionally, with the MPEG2 scheme widely used for coding moving picture information, coding of each picture frame is performed through any of intraframe coding, forward direction interframe coding and bidirectional interframe coding. Depending on which of those coding schemes is applied, the coded frame is referred to either as an I-picture, a P-picture or a B-picture.
With the MPEG2 scheme, a reference picture of a P-picture is limited to an immediately-preceding I-picture or P-picture. This allows only I-pictures and P-pictures to be extracted, and enables correct decoding and playback. As a result, for instance, fast playback is easily achieved.
However, the H.264 scheme not only allows a P-picture to use a plurality of pictures as a reference picture, but also allows a B-picture to be used as a reference picture. Therefore, when attempting playback by extracting only I-pictures and P-pictures from a coded image stream, it is difficult to perform normal decoding on a P-picture that uses a B-picture as a reference picture.
According to reference document D1, when performing fast playback, for an interval between an I-picture and a picture immediately preceding the next I-picture, a group of temporally-consecutive frames from the I-picture to at least a first P-picture is decoded. Subsequently, only I-pictures and P-pictures included in the decoded frame group are played back.
As described above, with the H. 264 scheme, as a result of allowing flexible selection of a reference picture during interframe coding, in principle, only I-pictures may be used for special playback.
Reference document D1 describes that an I-picture and a P-picture may be decoded if the pictures between the I-picture and the P-picture are removed. However, since the H.264/AVC standard allows a B-picture referenced by a P-picture to reference a picture that is even older than an I-picture, there is a risk that the P-picture may not be decodable. Measures for such cases are not disclosed in reference document D1.
While encoding for special playback mode may conceivably be performed by increasing the I-pictures, such encoding will reduce coding efficiency, and as a result, the feature of the H.264 scheme in that high-efficiency coding is realized by allowing flexible reference relationships may no longer be utilized.
In addition, the MPEG scheme manages a picture group consisting of a predetermined number of consecutive pictures as a group of pictures (GOP), and manages coding on a per-GOP basis. While it is desirable to increase the number of pictures configuring a GOP in order to achieve high-efficiency coding, reference document D1 does not provide a description on a special playback method in a case where a GOP structure changes during coding.